Fixations, regressive saccades, and parafoveal views: notes on the science of reading

Perhaps it’s an acquired taste, or just an eccentricity, but I rather like long passages of unbroken text. A well designed page, carefully set using a pleasing typeface, is a pleasure to contemplate.

But I’m aware I don’t treat a page with particular reverence when I’m actually reading it. I don’t read each each word in turn and move on, in orderly fashion, to the next sentence. My eyes skip back and forth across sentences, paragraphs and even pages as I work to grasp the text’s meaning, like a bee gathering pollen from a meadow.

I’ve never considered the how the mind processes information in this way. It turns out the science of reading has an austere poetry of its own.

The science of reading

In her useful little book Content Design, Sarah Richards, well known for her work in helping turn GOV.UK into one of the world’s most accessible websites, gives a good overview of what we do when we read – ‘the involuntary mechanics that govern how humans take in information’.

Psychologists refer to the process by which the eye lands on a word as ‘fixation’. It happens in an instant, but there are few microsteps.

The eye takes 100 milliseconds (ms) to identify a word, often just by recognising the first few characters. It then grasps its meaning by setting it in the context of the words that immediately follow, before looking ahead to the next fixation point (the ‘parafoveal view’).

The leap the eye makes when absorbing meaning is known as a ‘saccade’ (pronounced ‘sa-kaid’). Meaning is only grasped during the fixation, when the eyes are focused. As Richards notes, if you try to watch your eyes move in a mirror you can’t see the movement.

But during the saccade, as the eye moves to the next fixation the brain is working, making a narrative of all the words that have been encountered at previous fixation points. We don’t even have to read the words between the fixations. The mind infers what they are likely to be. Those learning a language need to read most of the words on a page. But an experienced reader can miss a third of them and still grasp the content.

All of this happens in milliseconds. The eye makes four or five saccades a second, and every couple of seconds moves back to words that have already been read but whose meaning has not yet been registered. These backwards steps are called ‘regressive saccades’.

We pass smoothly over or even just skip the words with which we are most familiar. But although most of us have a vocabulary of around 15,000 words, we need a little more time to digest those we don’t often encounter: an extra 100ms to be precise.

The science of writing

All of this provides a scientific basis for some well known principles of good writing.

First, it explains why it’s so important to get line lengths right. If lines are too long the eye can’t identify the fixation to which it needs to return: we lose our place when we’re trying to re-read something. If lines are too short the text doesn’t seem to flow enough to allow us to develop a comfortable reading rhythm. Regressive saccades will increase as the eye repeatedly returns to fixations. As typographers have known since Gutenburg, and as medieval and ancient manuscripts testify, for centuries before that, the ideal line length is around 40 to 60 characters.

Second, our ability to quickly grasp familiar words is why you should use short words most readers know well when you are trying to convey information efficiently. Your readers can then settle into a natural, flowing reading rhythm, their saccades skipping smoothly through the text.

Third, if content has a functional purpose – like information on a government website or detailing a product – it should be as easy to read as possible, presented to take the reader on a seamless journey through the content.

Nielsen Norman Group eyetracking screenshot

If it isn’t readers will resort to the improvised scanning noted in a series of famous articles by the web usability advocate Jakob Nielsen. In a highly influential little post written back in 1997, and still well worth reading, Nielsen’s research found that most users always scan web pages. Only one in five read them as the writer intended.

A few years later Nielsen wrote another celebrated article describing how his agency ran eyetracking tests showing that if a page is just an anonymous block of text, offering no visual cues as to how it should be digested, readers will resort to reading the pages in an F-shaped pattern.

The eyetracking heatmaps produced by the study showed two horizontal stripes followed by a vertical stripe. Readers usually start conventionally enough at the top left and scan the first line of text, a movement that forms the F’s top bar. They then move down, not necessarily to the second line, and leave another horizontal trail usually a little shorter than the first – the F’s lower bar. Finally, the eye moves down along the content’s left side, forming the F’s stem.

The value of longer copy

There is an important proviso. If the text is of intrinsic value to the reader, not merely an end to gaining information, they will be prepared to read it patiently. The text might be a feature article, essay or academic paper to which the reader comes prepared to invest time. I discuss the – sometimes suprising – value of longer copy in another post.

But when readers are just looking for information, and that is most of the time, it is important to structure the page correctly. Much has been written about that – I summarise my thoughts as to how it should be done here.

The image above is a detail from A Youthful Saint Reading by Girolamo de Santacroce.